Differential D6566
Cheaper sampling for distant mesh area lights Authored by Antonis Ryakiotakis (psy-fi) on Jan 12 2020, 8:44 PM.
Details
This patch uses Erik Heitz's A Low-Distortion Map Between Triangle and Square to skip square root evaluation in mesh light sampling. There may be more places in cycles where this could be useful, for now just using it in mesh lights. Based on ideas from https://pharr.org/matt/blog/2019/02/27/triangle-sampling-1.html
Diff Detail
Event TimelineComment Actions I'm testing this in a very simple file with an emitting suzanne mesh and a diffuse plane on the ground but I can't see any performance difference. Probably it's not worth submitting but will leave here just in case it's useful somehow. Comment Actions srqt() isn't that much of a performance issue any more as it once was. That said, the newer mapping should give us a better sample distribution. Comment Actions Fine with me, but ensure the Cycles regression tests pass and update the reference images if needed. |